| 528 | 2 | 395 |
| 下载次数 | 被引频次 | 阅读次数 |
死鱼对于水域生态和饮水安全存在巨大威胁，由于水面环境复杂，导致现有目标检测算法在死鱼检测任务中存在漏检、误检等情况。为此，以单次多边框检测(single shot multibox detector, SSD)为基础提出一种基于多尺度特征融合和注意力机制的水面死鱼检测方法FFA-SSD(SSD with feature fusion and attention)。首先，采用计算量和参数量更少且特征提取能力更强的残差网络ResNet50替换VGG16主干网络；其次，设计了多尺度特征融合模块，增强浅层特征和高层语义信息的融合；然后，引入通道注意力机制，抑制特征融合带来的冗余信息干扰，提升网络对目标的关注度；最后，设计了一种适用于小目标检测的数据增强算法，扩充训练数据中的小目标数量，丰富训练背景。实验结果表明，同现有目标检测算法相比，FFA-SSD算法可以更好地识别水面死鱼，检测精度达到93.5%。
Abstract：Dead fish could pose a huge threat to water ecology and safety of drinking water. With complex water surface environment, the existing object detection algorithms had some flaws such as missed and false detections in small target. Therefore, a dead fish detection method on water surface based on the multi-scale feature fusion and attention mechanism, SSD with feature fusion and attention(FFA-SSD) was proposed. Firstly, the residual network ResNet50 with less computation and fewer parameters and better feature extraction ability was used to replace the VGG16 backbone network. Then, a multi-scale feature fusion module was designed to enhance the fusion of shallow features and high-level semantic information. Finally, a channel attention mechanism was introduced to suppress the interference of redundant information brought by feature fusion and to improve the network′s focus on the target. In addition, a data enhancement algorithm applicable for small target detection was designed to increase the number of small targets in the training data and to enrich the training background. The experimental results showed that compared with other target detection algorithms, the recognition function of FFA-SSD algorithm for dead fish on the water surface was better, and the detection accuracy was at 93.5%.
[1] YU G Y,WANG L,HOU M X,et al.An adaptive dead fish detection approach using SSD-MobileNet[C]//2020 Chinese Automation Congress.Piscataway:IEEE Press,2021:1973-1979.
[2] 邓姗姗，黄慧，马燕.基于改进Faster R-CNN的小目标检测算法[J].计算机工程与科学，2023,45(5):869-877.DENG S S,HUANG H,MA Y.A small object detection algorithm based on improved Faster R-CNN[J].Computer engineering & science,2023,45(5):869-877.
[3] ZHENG J L,LIU Y.A study on small-scale ship detection based on attention mechanism[J].IEEE access,2022,10:77940-77949.
[4] 李国进，姚冬宜，艾矫燕，等.基于改进Faster R-CNN的水面漂浮物识别与定位[J].信阳师范学院学报(自然科学版),2021,34(2):292-299.LI G J,YAO D Y,AI J Y,et al.Detection and localization of floating objects via improved faster R-CNN[J].Journal of Xinyang normal university (natural science edition),2021,34(2):292-299.
[5] DUC MINH T,HOA N T N,LE T H.A model for floating garbage detection and quantification using fixed camera[C]//2022 9th NAFOSTED Conference on Information and Computer Science.Piscataway:IEEE Press,2022:389-393.
[6] CHENG Q Q,WANG H J,ZHU B,et al.A real-time UAV target detection algorithm based on edge computing[J].Drones,2023,7(2):95.
[7] HE X Q,WANG J C,CHEN C B,et al.Detection of the floating objects on the water surface based on improved YOLOv5[C]//2021 IEEE 2nd International Conference on Information Technology,Big Data and Artificial Intelligence (ICIBA).Piscataway:IEEE Press,2021:772-777.
[8] ZHANG L L,WEI Y X,WANG H B,et al.Real-time detection of river surface floating object based on improved RefineDet[J].IEEE access,2021,9:81147-81160.
[9] LI X L,TIAN M J,KONG S H,et al.A modified YOLOv3 detection method for vision-based water surface garbage capture robot[J].International journal of advanced robotic systems,2020,17(3):172988142093271.
[10] 张堡瑞，肖宇峰，郑又能.基于激光雷达与视觉融合的水面漂浮物检测[J].应用激光，2021,41(3):619-628.ZHANG B R,XIAO Y F,ZHENG Y N.Detection of floating objects on water surface based on fusion of lidar and vision[J].Applied laser,2021,41(3):619-628.
[11] CHENG Y W,XU H,LIU Y M.Robust small object detection on the water surface through fusion of camera and millimeter wave radar[C]//2021 IEEE/CVF International Conference on Computer Vision.Piscataway:IEEE Press,2021:15243-15252.
[12] 马天凤，杨震，罗勇，等.基于深度特征融合的红外弱小目标检测方法[J].郑州大学学报(理学版),2023,55(3):65-72.MA T F,YANG Z,LUO Y,et al.Infrared dim small target detection method based on depth feature fusion[J].Journal of Zhengzhou university (natural science edition),2023,55(3):65-72.
[13] 肖进胜，赵陶，周剑，等.基于上下文增强和特征提纯的小目标检测网络[J].计算机研究与发展，2023,60(2):465-474.XIAO J S,ZHAO T,ZHOU J,et al.Small target detection network based on context augmentation and feature refinement[J].Journal of computer research and development,2023,60(2):465-474.
[14] 陈科圻，朱志亮，邓小明，等.多尺度目标检测的深度学习研究综述[J].软件学报，2021,32(4):1201-1227.CHEN K Q,ZHU Z L,DENG X M,et al.Deep learning for multi-scale object detection:a survey[J].Journal of software,2021,32(4):1201-1227.
[15] RAVEENA S,SURENDRAN R.ResNet50-based classification of coffee cherry maturity using deep-CNN[C]//2023 5th International Conference on Smart Systems and Inventive Technology (ICSSIT).Piscataway:IEEE Press,2023:1275-1281.
[16] SONG T N,QIN W W,LIANG Z,et al.Research on CNN for anti-missile object detection algorithm based on improved attention mechanism[C]//2021 40th Chinese Control Conference.Piscataway:IEEE Press,2021:8286-8291.
[17] HOU Q S,XING J S.SSD object detection algorithm based on KL loss and Grad-CAM[J].Acta electronica sinica,2020,48(12):2409-2416.
基本信息:
DOI：10.13705/j.issn.1671-6841.2023149
中图分类号:TP18;TP391.41;X832
引用信息:
[1]杨帅鹏,李贺,刘金江,等.基于多尺度特征融合和注意力机制的水面死鱼检测方法[J].郑州大学学报(理学版),2024,56(06):32-38.DOI:10.13705/j.issn.1671-6841.2023149.
基金信息:
国家自然科学基金项目(62002180);; 河南省科技攻关项目(202102210362,232102210149);; 河南省高等学校重点科研项目(24A520030);; 南阳师范学院实验室开放项目(SYKF2021029)